| 1743 | """Preprocessor for Target Speaker Extraction.""" |
| 1744 | |
| 1745 | def __init__( |
| 1746 | self, |
| 1747 | train: bool, |
| 1748 | train_spk2enroll: Optional[str] = None, |
| 1749 | enroll_segment: int = None, |
| 1750 | load_spk_embedding: bool = False, |
| 1751 | load_all_speakers: bool = False, |
| 1752 | # inherited from EnhPreprocessor |
| 1753 | rir_scp: Optional[str] = None, |
| 1754 | rir_apply_prob: float = 1.0, |
| 1755 | noise_scp: Optional[str] = None, |
| 1756 | noise_apply_prob: float = 1.0, |
| 1757 | noise_db_range: str = "3_10", |
| 1758 | short_noise_thres: float = 0.5, |
| 1759 | speech_volume_normalize: float = None, |
| 1760 | speech_name: str = "speech_mix", |
| 1761 | speech_ref_name_prefix: str = "speech_ref", |
| 1762 | noise_ref_name_prefix: str = "noise_ref", |
| 1763 | dereverb_ref_name_prefix: str = "dereverb_ref", |
| 1764 | use_reverberant_ref: bool = False, |
| 1765 | num_spk: int = 1, |
| 1766 | num_noise_type: int = 1, |
| 1767 | sample_rate: int = 8000, |
| 1768 | force_single_channel: bool = False, |
| 1769 | channel_reordering: bool = False, |
| 1770 | categories: Optional[List] = None, |
| 1771 | data_aug_effects: List = None, |
| 1772 | data_aug_num: List[int] = [1, 1], |
| 1773 | data_aug_prob: float = 0.0, |
| 1774 | speech_segment: Optional[int] = None, |
| 1775 | avoid_allzero_segment: bool = True, |
| 1776 | flexible_numspk: bool = False, |
| 1777 | ): |
| 1778 | super().__init__( |
| 1779 | train, |
| 1780 | rir_scp=rir_scp, |
| 1781 | rir_apply_prob=rir_apply_prob, |
| 1782 | noise_scp=noise_scp, |
| 1783 | noise_apply_prob=noise_apply_prob, |
| 1784 | noise_db_range=noise_db_range, |
| 1785 | short_noise_thres=short_noise_thres, |
| 1786 | speech_volume_normalize=speech_volume_normalize, |
| 1787 | speech_name=speech_name, |
| 1788 | speech_ref_name_prefix=speech_ref_name_prefix, |
| 1789 | noise_ref_name_prefix=noise_ref_name_prefix, |
| 1790 | dereverb_ref_name_prefix=dereverb_ref_name_prefix, |
| 1791 | use_reverberant_ref=use_reverberant_ref, |
| 1792 | num_spk=num_spk, |
| 1793 | num_noise_type=num_noise_type, |
| 1794 | sample_rate=sample_rate, |
| 1795 | force_single_channel=force_single_channel, |
| 1796 | channel_reordering=channel_reordering, |
| 1797 | categories=categories, |
| 1798 | data_aug_effects=data_aug_effects, |
| 1799 | data_aug_num=data_aug_num, |
| 1800 | data_aug_prob=data_aug_prob, |
| 1801 | speech_segment=speech_segment, |
| 1802 | avoid_allzero_segment=avoid_allzero_segment, |